My bot is stupid.

[inner cradle]

My bot remembers nothing and I always have to reiterate, what am I doing wrong?

[hardy lynx]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[sour plinth]

Based on the FAQ and memory documentation, here's what could be causing the memory issues:

Common causes of "bot remembers nothing":

1. Sandboxed session with read-only workspace - If running in Docker/sandbox with workspaceAccess: "ro" or "none", memory files can't be written. Check your config for agents.defaults.sandbox.

2. Wrong workspace location - If the gateway restarts and uses a different workspace each time, memory files won't persist. Confirm agents.defaults.workspace points to a persistent directory.

3. No memory files exist yet - Clawdbot doesn't auto-create MEMORY.md. You need to:

   • Create MEMORY.md in your workspace for long-term memory
   • Create memory/ folder for daily notes

4. Group/thread context - Group threads don't share memory with DMs. If you're testing memory in a group, the bot might not load MEMORY.md there (by design for privacy).

5. Memory search disabled - Check if memorySearch.enabled is false in config, which would disable semantic memory recall.

Quick debug steps:

# Check where your workspace is
clawdbot config get agents.defaults.workspace

# Check if memory files exist
ls -la ~/clawd/memory/ 2>/dev/null || echo "No memory folder"
cat ~/clawd/MEMORY.md 2>/dev/null || echo "No MEMORY.md"

Enable verbose memory logging:
Add to config:

{
  logging: { level: "debug" },  // to see memory tool calls
  agents: {
    defaults: {

      memorySearch: { enabled: true }
    }
  }
}

The most likely culprit is either #1 (sandbox blocking writes) or #3 (no memory files created yet). Can you share your config or check if any of these apply? 🦐

[inner cradle]

Have there been any guides or tools to help deploy on railway?

[sour plinth]

Hey! 🦐 No dedicated Railway docs exist yet, but Railway is Docker-based so you can use the existing Docker guide as your base:

Key Railway-specific notes:

1. Use the Docker flow → Follow Docker install and Hetzner VPS guide for the pattern (same concept, different host)

2. Environment variables - Set these in Railway's dashboard:

   CLAWDBOT_GATEWAY_TOKEN=...
   ANTHROPIC_API_KEY=... (or setup-token)
   OPENAI_API_KEY=... (if needed)
   

3. Persistent storage - Railway has persistent volumes. Map:

   • /home/node/.clawdbot → persistent volume (config, auth, sessions)
   • /home/node/clawd → persistent volume (workspace, memory, skills)

4. Build & deploy - Railway auto-deploys from your Dockerfile. Use the Dockerfile pattern from the Hetzner guide (build binaries at image time, don't install at runtime).

5. Health check - Railway needs health checks. Add to your Dockerfile/service:

   HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
     CMD node dist/index.js health || exit 1
   

6. Webhooks - If using Discord/Telegram/Slack, configure their webhook URLs to point to your Railway app URL.

Quick starter railway.json:

{
  "build": {
    "builder": "DOCKERFILE",
    "dockerfilePath": "Dockerfile"